Increasing Prevalence and Severity of Clostridium difficile Colitis in Hospitalized Patients in the United States FREE

Clostridium difficile is a common cause of nosocomial diarrhea, with substantial associated morbidity and potential mortality. This pathogen, a gram-positive bacillus, inhabits the human intestinal tract in 1% to 3% of healthy adults and in about 20% of patients receiving antibiotics.¹ Yet, a subset of colonized patients becomes symptomatic when the usual colonic flora is altered and overtaken by this pathogen. Clinically, C difficile is associated with a spectrum of conditions, ranging from asymptomatic states to severe or complicated diarrheal illnesses that can ultimately lead to death. For patients with life-threatening C difficile colitis, surgical treatment, generally subtotal colectomy with end ileostomy, is the only effective therapy, but is associated with a high rate of morbidity and mortality.²

Three million new cases of C difficile colitis occur in the United States each year; as many as 10% of patients hospitalized for at least 2 days are affected.³ Anecdotal evidence and some case series suggest that C difficile colitis has become more common and potentially more pathogenic. An increased incidence of C difficile colitis was reported in hospitalized patients in Pittsburgh, Pennsylvania, and Portland, Oregon.^4- 5 Similarly, population-based data from Quebec revealed an increased incidence of C difficile colitis and an increasing associated mortality rate.⁶ In Quebec, an outbreak of C difficile colitis in 2003 is estimated to have been responsible for as many as 2000 deaths.⁷ The high mortality rate in that outbreak may have been the result of epidemic strains of C difficile with particularly high virulence and increased transmissibility.⁸

Given the potential changes in the epidemiological and biological activity of this frequent nosocomial disease, it is essential that patterns of this disease are evaluated. However, population-based data in the United States are lacking. We, therefore, designed this study, using population-based data, to determine the prevalence and severity (as measured by the mortality rate and the colectomy rate) of C difficile colitis in hospitalized patients in the United States during the past decade.

We obtained discharge data from the Nationwide Inpatient Sample (NIS) for January 1, 1993, through December 31, 2003, via the Healthcare Cost and Utilization Project of the Agency for Healthcare Research and Quality. The NIS—the largest source of all-payer hospital discharge information in the United States—is a unique and powerful tool. It includes data from about 7 million hospital stays per year in 1000 hospitals located in 35 states; thus, it approximates a 20% stratified sample of US community hospitals. It provides information on patient demographics, socioeconomic factors, admission profiles, hospital profiles, state codes, discharge diagnoses, procedure codes, total charges, and vital status at hospital discharge. Along with other hospital discharge databases, the NIS has been used extensively in the past to review trends in surgical care and outcomes,⁹ volume outcome relationships,¹⁰ and disparities in care.¹¹ A data use agreement is held by the Agency for Healthcare Research and Quality; for our study, all study protocols were considered exempt by the University of Minnesota institutional review board.

We used diagnostic codes from the International Classification of Diseases, Ninth Revision (ICD-9) to identify all patients who were discharged with a diagnosis of C difficile colitis (ICD-9 code 8.45). The diagnostic code for C difficile colitis was first introduced in October 1992, so we excluded data from before 1993.

We evaluated 2 groups of patients with C difficile colitis: (1) those with a principal diagnosis of C difficile colitis (a group of patients admitted primarily for C difficile colitis) and (2) those with a principal or secondary diagnosis of C difficile colitis (a group for whom C difficile colitis contributed to the hospital stay but may not have been the primary reason for admission). Patients with a secondary diagnosis of C difficile colitis may have had multiple other significant conditions contributing to morbidity or mortality. In an attempt to ascertain the effect of C difficile colitis on mortality, without the effect of other more potentially serious conditions that are associated with high mortality, we performed analyses for all outcome measures in patients with C difficile colitis as a principal diagnosis and in patients with any diagnosis of C difficile colitis.

We determined the prevalence of disease by recording all ICD-9 codes for C difficile colitis as a principal or secondary diagnosis and standardizing with a denominator that included the total discharges for that year of diagnosis. Thus, the prevalence was described in a population of 100 000 discharged patients and presented as the C difficile discharge rate. This represents the frequency of hospitalization for C difficile colitis (principal diagnosis) and the frequency of C difficile colitis influencing hospital stay (any diagnosis) in the United States.

We recorded in-hospital deaths for all patients with a diagnosis of C difficile colitis. We calculated case fatality (as a proportionate variable) by dividing the number of deaths in patients with C difficile colitis by the total number of patients with the disease. To ascertain the burden of C difficile colitis on mortality, we also calculated the total mortality rate (described in a population of 100 000 discharged patients) by dividing the number of deaths of patients with C difficile colitis by the total number of patient discharges. We determined case fatality and the total mortality rate for patients with a principal diagnosis of C difficile colitis and, separately, for patients with any diagnosis of C difficile colitis.

We used standard ICD-9 codes to select patients with a diagnosis of C difficile colitis who underwent subtotal colectomy (ICD-9 code 45.8). We excluded patients who underwent segmental colectomy, because the surgical standard of care for complicated C difficile colitis is subtotal colectomy with end ileostomy.¹² In addition, we excluded patients with a diagnosis of diverticulitis, diverticular hemorrhage, and colon cancers. We determined the colectomy rate for patients with a principal diagnosis of C difficile colitis and, separately, for patients with any diagnosis of C difficile colitis.

Hospital stay was defined as the difference in days between the admission and discharge dates. Length of stay was coded as 0 for patients discharged during the day of admission. Because immediate care stays have been declining and severity indexes for hospitalized patients have been increasing, we adjusted for comorbidity using the Deyo modification of the Charlson comorbidity index.¹³ Briefly, we ascertained the presence of 17 comorbid conditions and then weighted them according to the original report by Charlson et al.¹⁴ An elevated Charlson comorbidity index has been shown to correlate with mortality.¹⁴

We compared demographic variables over 3 periods (January 1, 1993–December 31, 1996, January 1, 1997–December 31, 2000, and January 1, 2001–December 31, 2003) using the χ² test for categorical variables and the t test for continuous variables. For each of the 11 years (1993-2003), we calculated the prevalence, case fatality, total mortality rate, and colectomy rate for patients with a principal diagnosis of C difficile colitis and, separately, for those with any diagnosis of C difficile colitis. Using the Cochran-Armitage trend test on 1 df, we determined whether rates changed over time. To control for potential confounders, we constructed a logistic regression model to evaluate the relationship between the year of diagnosis (continuous variable) and the prevalence of C difficile colitis, case fatality, and colectomy rate; we adjusted for age (continuous variable), sex, race (white, black, Hispanic, other, or missing), payer status (Medicare, Medicaid, private payer, or other), and Deyo comorbidity score (continuous variable). Because the NIS removes race status for about 20% of all patients, we performed the regression twice (once with missing race as a separate variable and once after excluding all patients with missing race).

All statistical analyses were performed using SAS statistical software, version 9.13 (SAS Institute Inc, Cary, North Carolina). All tests of statistical significance were 2-sided; P<.05 was considered statistically significant. Because the NIS is a stratified probability sample of US community hospitals, we adjusted our calculations for survey sampling characteristics (probability weights and stratification).

The NIS contains 78 091 119 discharge abstracts over 11 years. During this period, 299 453 patients were discharged with a principal or secondary diagnosis of C difficile colitis. The prevalence of disease in the 11-year study period was 383 cases per 100 000 discharged patients. The rate of C difficile colitis discharges increased from 261 cases per 100 000 discharged patients in 1993 to 546 cases per 100 000 discharged patients in 2003 (Figure 1) (P < .001), a 109% increase. Of all patients with a C difficile colitis diagnosis, 69 373 (23.2%) had a principal diagnosis of C difficile colitis. Patients with a principal diagnosis of C difficile colitis had fewer other secondary diagnoses and a significantly lower Deyo comorbidity score (mean ± SEM, 1.23 ± 0.004) compared with patients with any diagnosis of C difficile colitis (mean ± SEM, 1.82 ± 0.007) (P < .001). The rate of principal diagnosis of C difficile colitis during the 11-year study period was 88.8 cases per 100 000 discharged patients. The principal diagnosis of C difficile colitis increased from 71.4 cases per 100 000 discharges in 1993 to 127.0 cases per 100 000 discharges in 2003 (P < .001), a 77.8% increase.

Proportionately, more patients with C difficile colitis were women than men (Table 1), although this difference decreased over time (59.80% of cases of C difficile colitis in women during the earlier period [January 1, 1993–December 31, 1996] vs 58.90% of cases in the later period [January 1, 2001–December 31, 2003]). During the 11-year study period, the mean ± SEM age increased from 65.60 ± 0.09 years in the earlier period (January 1, 1993–December 31, 1996) to 67.60 ± 0.06 years in the later period (January 1, 2001–December 31, 2003). Payer data revealed that many patients used Medicare insurance; this proportion increased from 66.10% in the earlier period to 67.70% in the later period. The mean ± SEM Deyo comorbidity score was 1.82 ± 0.004; it increased from 1.75 ± 0.009 in the earlier period to 1.88 ± 0.007 in the middle period (January 1, 1997–December 31, 2000), and then declined to 1.81 ± 0.006 in the later period. In addition, the mean ± SEM length of stay was 14.10 ± 0.04 days; it decreased from 15.90 ± 0.09 days in the earlier period to 13.50 ± 0.04 days in the later period (Table 1).

Abdominal colectomy was performed in 0.27% of patients with any diagnosis of C difficile colitis (ie, 2.7 colectomies per 1000 patients hospitalized with either a principal or a secondary diagnosis of C difficile colitis). The colectomy rate increased from 1.2 colectomies per 1000 patients in 1993 to 3.4 colectomies per 1000 patients in 2003 (Table 2). In the group of patients with a principal diagnosis of C difficile colitis, the colectomy rate averaged 0.41% (ie, 4.1 cases per 1000 patients with C difficile colitis); it ranged from 1.8 cases per 1000 patients in 1993 to 6.4 cases per 1000 patients in 2003 (P < .001).

Of the 299 453 patients hospitalized with a principal or secondary diagnosis of C difficile colitis, 26 246 (8.76%) died. The case fatality increased from 7.84% in 1993 to 9.26% in 2003 (P < .001) (Figure 2). Of the patients with a principal diagnosis of C difficile colitis, 2377 died (ie, case fatality of 3.43%; range, 3.26% in 1993 to 3.85% in 2003) (P < .001).

The total mortality rate for the 11-year study period in patients with C difficile colitis was 33.6 deaths per 100 000 discharged patients. The total mortality rate increased from 20.3 deaths per 100 000 discharges in 1993 to 50.2 deaths per 100 000 discharges in 2003 (Table 3), a 147% increase. Of the patients with a principal diagnosis of C difficile colitis, the total mortality rate averaged 3.04 deaths per 100 000 discharged patients; it ranged from 2.32 deaths per 100 000 discharged patients in 1993 to 4.26 deaths per 100 000 discharged patients in 2003 (P < .001), an 83.6% increase.

We evaluated the influence of age, sex, race, payer status, and Deyo comorbidity score on prevalence of C difficile colitis, case fatality, total mortality rate, and colectomy rate (Table 2 and Table 3). Although the prevalence and the total mortality rate increased with the patients' increasing age, the colectomy rate declined at the higher extreme of age (ie, ≥70 years). Although proportionally more patients with C difficile colitis were women, men had a higher case fatality and a higher colectomy rate. In addition, patients with private insurance had a lower overall rate of C difficile colitis, but a relatively high colectomy rate. As the number of comorbidities increased, so did the prevalence, case fatality, and total mortality rate; however, the colectomy rate declined (Table 2 and Table 3).

In our multivariate analysis, the year of diagnosis remained a strong predictor of the prevalence of C difficile colitis, case fatality, total mortality rate, and colectomy rate after we controlled for potential confounders, including age, sex, race, payer status, and Deyo comorbidity score (Table 4). After adjusting for covariates, we found that the diagnosis of C difficile colitis was significantly more likely with increasing calendar year (odds ratio [OR], 1.05; 95% confidence interval [CI], 1.05-1.05; P < .001). Also, patients with C difficile colitis were more likely to die with increasing calendar year, and were more likely to undergo colectomy. Similarly, with each passing calendar year, patients were more likely to have a primary diagnosis of C difficile colitis (OR, 1.07; 95% CI, 1.06-1.07; P < .001); also, patients with a primary diagnosis of C difficile colitis were more likely to die (OR, 1.01; 95% CI, 1.00-1.03; P < .05), and were more likely to undergo colectomy (OR, 1.09; 95% CI, 1.05-1.12; P < .001).

The evidence is substantial that the epidemiological features of C difficile are changing; increases in disease prevalence have been demonstrated in population-based studies in Canada and Sweden.^{4- 6,15- 17} In the United States, from 1987 through 2001, nosocomial rates of reported C difficile infection increased in patients admitted to intensive care units of large hospitals¹⁸ and increased hospitalwide in patients admitted to small hospitals. But, until now, recent population-based US data have been lacking. Our US-based study of almost 300 000 hospitalized patients with a diagnosis of C difficile colitis provides compelling additional evidence that the burden of illness due to this disease has increased. During our 11-year study period, we found an increasing prevalence of C difficile colitis in patients admitted to US hospitals. Because we also identified an increase in the severity of C difficile colitis (as measured by increasing case fatality and an increasing colectomy rate), the increase in prevalence is not merely because of an increased ascertainment of C difficile colitis cases.

From a stratified sample of all US hospital patient discharges, we demonstrated an increased rate of C difficile diagnoses, elevated case fatality, an increased total mortality rate, and an increased colectomy rate. Hospital discharge with a C difficile diagnosis was significantly more likely with increasing calendar year. In addition, the likelihood of death and of treatment with colectomy also significantly increased over time. Case fatality for the group with a principal diagnosis of C difficile colitis also increased, but that increase was more modest. Patients with a principal diagnosis of C difficile colitis had a lower mean comorbidity score compared with patients with any diagnosis of C difficile colitis. The principal diagnosis, defined as the diagnosis that is established to be most responsible for causing the patient's admission to the hospital, is subject to the interpretation of hospital coders. Critically ill patients with C difficile colitis could be coded with C difficile colitis as a principal diagnosis, or could be coded using a different principal diagnosis and diagnosis-related groups, such as septicemia or shock, which are subject to a higher reimbursement schedule, with C difficile colitis as a secondary diagnosis. This may explain the higher case fatality in those with any C difficile colitis (compared with principal diagnosis only) and the proportionately greater increase in case fatality over time for that group.

Overall, the population-based data in our study provide supporting evidence for the clinical impression that the nature of C difficile colitis in hospitalized patients is changing,¹⁹ but did not allow determination of the cause of the change. A recent epidemic of C difficile colitis in Quebec was attributed to a particular strain of C difficile: toxinotype III, North American PFGE type 1, and PCR ribotype 027 (NAP1/027). This strain carries the binary toxin gene cdtB (cytolethal distending toxin B gene) and an 18–base pair deletion in tcdC; it hyperproduces toxins A and B.^6,8,20- 21 In addition, this hypervirulent strain is also associated with increased disease severity^6,21 and possibly transmissibility and has caused outbreaks in England,²² Canada,²³ and the United States.^8,24

Other factors may also be contributing to the increased frequency and severity of C difficile colitis. Evidence points to the increasing resistance of C difficile to standard first-line therapy.¹⁹ In the past, high response rates (90%) and low recurrence rates with the use of first-line metronidazole were found, but more recent reports^21,25 showed that the overall effectiveness of metronidazole is lower, particularly in intensive care unit patients.²⁶ A study²⁷ from the United Kingdom revealed a high incidence of resistant strains of C difficile across different geographic regions. The increased resistance is proposed to have occurred through acquisition of locally occurring mechanisms of resistance, thereby reducing treatment efficacy.²⁷ The increased use of fluoroquinolones, particularly those with increased anaerobic activity and gut excretion, has been associated with increased rates of C difficile by a number of researchers.^{22,25- 26,28- 29} In a retrospective cohort study of 5619 patients consecutively admitted to a single hospital, fluoroquinolones were most strongly associated with the development of C difficile (attributable hazard ratio, 3.4). The use of other antibiotics (eg, cephalosporins, macrolides, and clindamycin) was associated with a still clinically significant, but lower, risk of developing C difficile (attributable hazard ratio, 1.56-1.89).²¹

However, patients who are sicker (eg, those who are immunocompromised, those who are older, those who have low serum albumin levels, those who require an intensive care unit stay, or those who have undergone recent abdominal surgery) have an increased risk of severe C difficile colitis.^26,30- 32 Therefore, the increasing severity of C difficile colitis identified in our study may be because of the increasing acuity of illness of hospitalized patients in the United States. We did find an overall increase in the comorbidities of admitted patients during our study period, but even after adjusting for comorbidities, we still found an increase over time in the prevalence, case fatality, and colectomy rate for C difficile colitis. Similarly, we found an increase in the prevalence and in case fatality for all age groups. In our study, many of the deaths in patients with C difficile colitis (as the principal diagnosis and as any diagnosis) were because of underlying medical conditions, rather than direct C difficile infection. Nevertheless, C difficile colitis would be a complicating diagnosis in these patients with acute illness, potentially contributing to death. Our analysis indicated that the severity of C difficile colitis increased significantly, whether we evaluated it as the principal diagnosis or as any diagnosis contributing to hospital stay. Patients with a principal diagnosis of C difficile colitis also had a high colectomy rate and high case fatality, despite lower comorbidity scores.

Our study used population-based data with only limited information on patient and treatment factors, thereby limiting our evaluation of medical treatments such as antibiotic use. Also, our study did not permit us to obtain a denominator detailing the total number of C difficile toxin studies performed or to verify the presence of the disease in the coded cases, leading us to rely on the individual clinicians' judgment. The inability to verify actual cases or ascertain the proportion of toxin assays performed is a limitation, but the ICD-9 code for C difficile colitis is specific for pseudomembranous colitis and is not limited to toxin presence. For this reason, the increase in prevalence of C difficile colitis is likely an increase in clinically significant infection. In addition, the population we reviewed included many elderly Medicare patients, who had a high rate of C difficile colitis and higher case fatality. Although private payer patients were younger and had lower C difficile colitis rates and lower case fatality, our study was specifically aimed to provide population-based estimates of C difficile colitis, which ultimately affects the more elderly and debilitated patient. Given these limitations, the study has numerous strengths, including the quality data tracking system used by the NIS and the large sample size. The relative inclusivity of patients with varied diagnoses³³ is another advantage of the NIS, which is optimally constructed to evaluate patterns and trends of diseases in hospitalized patients across the United States.

Because the results of toxin assays were not available in our data set, we were not able to confirm the diagnosis of C difficile colitis. It is possible that the ability to diagnose mild cases of C difficile colitis may have improved over time with the use of newer toxin assays. Yet, any such increased ascertainment of disease would result in a lower severity of disease. In contrast, we found an increase in the severity of disease, as measured by case fatality and the colectomy rate, indicating that increased ascertainment could not explain our findings. We found that misdiagnosis, although possible, was highly unlikely, especially given our consistent findings across various ages, geographic regions, and comorbidities. More important, the NIS data that we used come from a number of practice settings across the United States, reflecting national trends rather than local phenomena.

In conclusion, our study of almost 300 000 hospitalized patients with C difficile colitis in the United States demonstrated that this disease has increased in prevalence and severity. To our knowledge, this study is the first population-based data analysis in the United States to demonstrate this change and confirms similar trends in C difficile colitis severity and prevalence found in other countries. While these data cannot determine the cause, further research in this area is clearly needed. Heightened awareness of the increasing disease burden of C difficile colitis is an important first step in controlling the public health ramifications of this important and morbid nosocomial infection.

Correspondence: Rocco Ricciardi, MD, MPH, Department of Colon and Rectal Surgery, Lahey Clinic, 41 Mall Rd, Burlington, MA 01805.

Accepted for Publication: March 13, 2006.

Author Contributions: Dr Ricciardi had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Study concept and design: Ricciardi, Rothenberger, Madoff, and Baxter. Acquisition of data: Ricciardi. Analysis and interpretation of data: Ricciardi and Baxter. Drafting of the manuscript: Ricciardi. Critical revision of the manuscript for important intellectual content: Ricciardi, Rothenberger, Madoff, and Baxter. Statistical analysis: Ricciardi and Baxter. Obtained funding: Ricciardi and Rothenberger. Study supervision: Rothenberger, Madoff, and Baxter.

Financial Disclosure: None reported.

Funding/Support: This study was supported by the University of Minnesota Academic Health Center's Clinical Scholars Research Grant.